Predicting food-web structure with metacommunity models

Background/Question/Methods

Food-web structure and dynamics play important roles in maintaining species diversity and functioning of ecosystems. Increasingly, regional-scale factors that drive spatial dynamics are being recognized as important determinants of local food-web structure. The metacommunity concept provides a framework for understanding the relative influence of local and regional-scale processes in generating diversity patterns across the landscape. Metacommunity models and empirical studies have focused mostly on assemblages of organisms within a single trophic level whose primary interactions are competitive. Studies of multi-trophic communities are predominantly restricted to simplified trophic motifs and rarely consider entire food webs. We tested the ability of the patch-dynamics, species-sorting, mass-effects, and neutral metacommunity models, as well as three hybrid models, to predict empirical patterns of food web structure and composition in the aquatic food web found in the northern pitcher plant, Sarracenia purpurea. We used empirical data to determine regional species pools and estimate dispersal probabilities, simulated local food-web dynamics, dispersed species from regional pools into local food webs based on the assumptions of each metacommunity perspective, and tested the relative fit of each metacommunity model to empirical data on food-web structure. 

Results/Conclusions

The species-sorting and patch-dynamics models most accurately predicted seven food web properties, suggesting that local-scale interactions were important in structuring Sarracenia food webs. A well-known example of an important local-scale interaction in the Sarracenia system is keystone predation by the mosquito, Wyeomyia smithii, which has been shown to alter species richness and composition in the Sarracenia food web. However, differences in dispersal abilities were also important in models that accurately reproduced empirical food web properties. All of the metacommunity models, even the best-fit ones, were unable to reproduce variation in species richness and trophic diversity.  This, coupled with the consistent underestimation of β-diversity by the majority of the metacommunity models, suggests that simulated food webs are more similar in species richness and composition than observed within the empirical metacommunities. We suggest that better quantifying dispersal and allowing different species within the same metacommunity to adhere to different metacommunity perspectives may increase the accuracy of metacommunty models in predicting food web structure. Although the models were tested using pitcher-plant food webs, the approach we have developed could be applied to any well-resolved food web for which data are available from multiple locations.